Developer guide

How to calculate AI cost per user in a mobile app

Learn how to calculate inference cost per active user, compare usage to subscription revenue and design practical AI usage limits.

Start with the real usage unit

Record provider, model, token usage, request count, time and the authenticated user or installation. Aggregate total AI cost per user over a consistent billing period.

Compare cost to revenue

For a subscription app, estimate contribution as net attributable subscription revenue minus inference cost and other variable expenses. Revenue should reflect platform fees and refunds; don't compare gross sales against net usage costs.

Why averages hide outliers

Averages are useful for budgeting but high-frequency users can consume most of your inference budget. Inspect the distribution: typical users, heavy users, premium subscribers and free accounts.

Enforce rules close to the request

Use plan-specific quotas, time windows and spending caps. Decide what happens when a limit is reached: hard block, graceful downgrade or a user-facing upgrade prompt.

Relaykit's approach

Relaykit is exploring the link between mobile app identity, user entitlements, AI requests and cost attribution. Early-access developers can help validate these workflows.

Building AI features for mobile?

Relaykit is inviting developers to shape early access. Free during early access; future paid pricing has not been finalized.

Request early access